Closed Bug 662417 Opened 9 years ago Closed 7 years ago

Install pulseaudio-libs-devel on Linux build machines

Categories

(Release Engineering :: General, defect, P3)

All
Linux
defect

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: kinetik, Assigned: jhopkins)

References

Details

(Whiteboard: [puppet][deployment])

Attachments

(1 file, 3 obsolete files)

The audio rewrite in bug 623444 is almost ready.  It introduces a build requirement for pulseaudio-libs-devel, so please arrange for this to be installed on the build machines.

Also, it'd be awesome if bug 574190 (update pulseaudio on the test machines) was looked at soon too.
Priority: -- → P3
See Also: → 574190
Whiteboard: [puppet][deployment]
Is a time set for completing this work?
Not yet. Is pulseaudio-libs-devel-0.9.10-1.el5.3.i386.rpm acceptable ? That's what yumdownloader finds.
That's linux32 bit, obviously. You'd want linux64 too, right ?
There are a bunch of dependencies too. Here's the output if you try install that rpm:
error: Failed dependencies:
	libpulse-browse.so.0 is needed by pulseaudio-libs-devel-0.9.10-1.el5.3.i386
	libpulse-mainloop-glib.so.0 is needed by pulseaudio-libs-devel-0.9.10-1.el5.3.i386
	libpulse-simple.so.0 is needed by pulseaudio-libs-devel-0.9.10-1.el5.3.i386
	libpulse.so.0 is needed by pulseaudio-libs-devel-0.9.10-1.el5.3.i386
	pulseaudio-libs = 0.9.10-1.el5.3 is needed by pulseaudio-libs-devel-0.9.10-1.el5.3.i386
	pulseaudio-libs-glib2 = 0.9.10-1.el5.3 is needed by pulseaudio-libs-devel-0.9.10-1.el5.3.i386
	pulseaudio-libs-zeroconf = 0.9.10-1.el5.3 is needed by pulseaudio-libs-devel-0.9.10-1.el5.3.i386
0.9.10 is seriously ancient (March 2008).  I just tried to build the code against that and it fails.  It looks like the minimum version needed will be 0.9.16.  Is that going to be possible?

It would be possible to make the code build against 0.9.10, but we'll miss out on a reasonably important optimization provided by the newer API.

Yeah, both x86 and x86_64 please.
jhford, can you take this bug ?
Ping?
(In reply to comment #6)
> jhford, can you take this bug ?

(In reply to comment #7)
> Ping?

I am going to be working on creating a new linux reference image which should deal with these issues.  Bugs will be filed today.
Ping? Any chance of an ETA?
This is going to addressed by bug 670707
Depends on: 670707
Any updates?
Ping.
Blocks: 617852
Blocks: 752401
Blocks: 751030
Depends on: 772446
Matthew, can you please confirm that pulseaudio-libs-devel :
- should only be installed on CentOS 6 build slaves (eg. not CentOS 5)
- should not be installed on -release or -esr* branches (or will it "do no harm" if installed for all branches, which would result in a simpler releng build config)
Flags: needinfo?(kinetik)
Blocks: 670707
No longer depends on: 670707
I'm reasonably confident that it's safe to install pulseaudio-libs-devel everywhere.  The only two pieces of code in our tree that use PulseAudio are libcubeb and webrtc (there's also code present in libsydneyaudio, but no build-time support for it), and both of them require explicit configuration (--enable-pulseaudio configure option for libcubeb, and use_pulseaudio setting in media.gyp for webrtc) to start using PA, so their build configuration will not change due to the PA headers being present.
Flags: needinfo?(kinetik)
Patch needs testing.

bhearsum: Any thoughts on whether we really want to add this package across the board or have it 'ride the trains'?

kinetic: Is this package also needed for b2g and fennec builds?
Attachment #705393 - Flags: feedback?(bhearsum)
Flags: needinfo?(kinetik)
Patch need testing.
Attachment #705395 - Flags: feedback?(bhearsum)
(In reply to John Hopkins (:jhopkins) from comment #15)
> kinetic: Is this package also needed for b2g and fennec builds?

We use different audio APIs on these platforms, so it's not needed.
Flags: needinfo?(kinetik)
(In reply to John Hopkins (:jhopkins) from comment #15)
> Created attachment 705393 [details] [diff] [review]
> [buildbot-configs] add pulseaudio-libs-devel
> 
> Patch needs testing.
> 
> bhearsum: Any thoughts on whether we really want to add this package across
> the board or have it 'ride the trains'?

We should ride the trains unless there's a strong reason not to.
Comment on attachment 705395 [details] [diff] [review]
[mozharness] add pulseaudio-libs-devel

Review of attachment 705395 [details] [diff] [review]:
-----------------------------------------------------------------

f+ for the mozilla-central parts. Per my previous comments, we should probably ride the trains.
Attachment #705395 - Flags: feedback?(bhearsum) → feedback+
Comment on attachment 705393 [details] [diff] [review]
[buildbot-configs] add pulseaudio-libs-devel

Review of attachment 705393 [details] [diff] [review]:
-----------------------------------------------------------------

f+, but needs to make sure pulseaudio isn't installed for aurora and up.
Attachment #705393 - Flags: feedback?(bhearsum) → feedback+
dump_masters output looks good.  Currently running some test builds in staging with this patch.
Attachment #705393 - Attachment is obsolete: true
Attachment #705395 - Attachment is obsolete: true
Attachment #706470 - Flags: review?(bhearsum)
Assignee: nobody → jhopkins
Comment on attachment 706470 [details] [diff] [review]
[buildbot-configs] add pulseaudio-libs-devel to inbound, central, projects

Review of attachment 706470 [details] [diff] [review]:
-----------------------------------------------------------------

::: mozilla/config.py
@@ +1302,5 @@
>                      value = deepcopy(value)
> +                #MERGE DAY - pulseaudio-libs-devel package rides the trains (bug 662417)
> +                if key == 'mock_packages':
> +                    if branch in ['mozilla-aurora', 'mozilla-beta', 'mozilla-release', 'mozilla-esr10', 'mozilla-esr17', 'mozilla-b2g18']:
> +                        value = [x for x in value if x != 'pulseaudio-libs-devel']

Can you move this disabling down with the other branch-specific overrides? Eg: https://github.com/mozilla/build-buildbot-configs/blob/1b197637ea7c3a467fd70eaf7ecb06176d120cfc/mozilla/config.py#L1547

This loop isn't a place that we do this sort of thing. Same for thunderbird but looks fine otherwise.
Attachment #706470 - Flags: review?(bhearsum) → review-
updated patch that moves exceptions to the end of the file.  dump_master output is identical to the previous patch.
Attachment #706470 - Attachment is obsolete: true
Attachment #707060 - Flags: review?(bhearsum)
Attachment #707060 - Flags: review?(bhearsum) → review+
Comment on attachment 707060 [details] [diff] [review]
[buildbot-configs] address bhearsum's review feedback

Landed in https://hg.mozilla.org/build/buildbot-configs/rev/b6487918b321
Attachment #707060 - Flags: checked-in+
In production.
Status: NEW → RESOLVED
Closed: 7 years ago
Resolution: --- → FIXED
No longer blocks: 617852
Blocks: 842863
Blocks: 842864
Product: mozilla.org → Release Engineering
You need to log in before you can comment on or make changes to this bug.